Texas sues GM for unlaw­ful­ly collecting and selling dri­vers' pri­vate data [pdf]

That's a cute tactic.

Claim that $BAD_SIDE_EFFECT would be bad, even if it's better than the status quo.

Nice way to prevent change.

out of curiosity: how much are you guys paying per month? i had, at that moment, 7 years of no accidents and the 9/1999 version of [1] and was at ~25euro/month

[1] https://www.adac.de/rund-ums-fahrzeug/auto-kaufen-verkaufen/...

> even on average

>> does this have any meaning at all

>> random chance, tend not to repeat

I think we agree about the nature of accidents -- they have some chance of occurring, the damages are often more than any one person can afford, and beyond a probabilistic assessment you have no way of knowing exactly how many wrecks will happen or what the damages will be.

My point is that you _can_ analyze those accidents probabilistically. Somebody who practices defensive driving, never drives over 25mph in a residential, only drives in broad daylight, and only travels 500 miles per year will have a very different baseline, both in number and severity of accidents, than somebody who habitually blows through residential stop signs at 60+mph and drives 50,000 miles per year.

Insurance concerns itself with flattening those spiky probability distributions. The first person will likely never severely injure a person even if they get in an accident, just from the difference in miles driven they're going to be in 100x fewer accidents per year, and probably much better than that because of their other safety practices (call it an additional 2x factor).

Just to have some hypothetical numbers to play with, the safer driver has an average of 0.001 wrecks per year, and the average damages might be $20k (fender benders, minor hospitalization, ...), so they have about $20/yr worth of risk to insure against (yes, I know you have to integrate over probability of different types of wrecks or whatever; this is a simple HN comment with ballpark numbers). The dangerous driver has 0.2 wrecks per year, and the average damages might be $100k (total both vehicles, major hospitalization, ...), so they have $20,000/yr to insure against.

_Insurance_ concerns itself with factoring in those relative risk profiles (along with the time value of money and whatnot; it gets a bit complicated) and guaranteeing that the individuals only have to pay their flat premiums (and optionally a flat deductible per wreck, though a decently high deductible is a good idea for most people) instead of risking bankruptcy and then some (it's the "and then some" that made car insurance mandatory in the US -- ensuring that the unlucky person you crashed into can still be made as financially whole as possible).

And that's all I meant with the "even on average" comment. If the safe driver's premiums were a bit more than $20/yr, and the dangerous driver's premiums were a bit more than $20,000/yr, that would be _fine_ from an insurance perspective.

The person I was replying to was talking about the "social function" of insurance, and my real point is that if the social function is to allow that dangerous, more expensive driver to continue to drive (not necessarily as bad of an idea as it sounds in the abstract -- if they couldn't legally drive, would they do something more dangerous or less insured instead? is it worth the additional costs we inflict on people when driving is nearly mandatory but we just won't let them?), _insurance_ isn't the tool to enable them. We should be thoughtful and explicit when providing those sorts of subsidies, instead of hiding and burying them in a tangentially related financial instrument.

Mind you, going 1-2 comments up the chain and referring to the data collection, I still think that's bad for other reasons, and using "too complicated" of models (under the assumption that they'll likely never be analyzed by a real person and have a chance of being egregiously wrong) isn't great either, but pricing insurance based on what you know about a person isn't bad in and of itself.

Mandatory car insurance, in theory, moves money from drivers (proportional to the risk they represent) to victims (proportional to the damages caused), plus a cut for the insurance company.

I'm missing something in your flow of ideas. Is it a welfare benefit because that idealized theory of insurance is itself a welfare benefit or because when insurance is mandated it doesn't live up to expectations?

So here's the problem. There are some crimes that have a relatively low probability of detection. It doesn't even have to be that low; say it's 40%.

Then there are people with nothing to lose. Some mid-level manager who isn't very good at it, isn't making his numbers. If he plays by the rules, he's about to get canned and lose his house and his wife. If he cheats and gets caught, there's a 40% chance he goes to jail, but a 60% chance he not only makes a lot of money but gets promoted to the head of the company because his numbers are so good. There are going to be too many people willing to take the risk in that case. It's going to happen a lot of the time.

To fix this you need to attach a penalty to someone who both has something to lose and is in a position to detect the crime. But the someone with something to lose are the owners and when they're diffuse passive investors they're not in a position to detect the crime.